Overview
Developed during the Google Cloud Platform (GCP) Hackathon 2023, this system delivers an intelligent time-series forecasting application designed to tackle post-pandemic hybrid workplace challenges. By accurately forecasting office attendance, cafeteria food catering demand, and hot-desk utilization, organizations can optimize facilities operational expenditure while enhancing employee workplace satisfaction.
Key Architectural Highlights
- Time-Series ML Forecasting Engine: Employs Facebook Prophet models (
prophet_model.json) trained on historical attendance matrices, seasonal holiday calendars, and day-of-week trends to project forward workplace resource requirements. - Google Cloud Platform Integration: Utilizes GCP infrastructure for scalable batch data ingestion, containerized execution via custom Docker containers, and robust pipeline automation.
- Interactive Streamlit Dashboard: Provides a responsive, real-time command interface (
app.py) allowing facility directors and floor managers to simulate sudden attendance surges, adjust prediction windows, and export demand schedules. - Exploratory Data Analytics (EDA): Includes comprehensive Jupyter analysis notebooks (
timeseries.ipynb,eda.ipynb) detailing feature selection, autocorrelation checks, and cross-validation metrics.
Technology Stack
- Machine Learning: Facebook Prophet, Pandas, NumPy, Scikit-Learn
- User Interface: Streamlit, Matplotlib, Seaborn
- Cloud Infrastructure: Google Cloud Platform (GCP), Docker
- Repository: GitHub - hackathon2023